Comparison of Mechanical Properties in Ultrafine Grained Commercial-Purity Aluminum (A1050) Processed by Accumulative Roll Bonding (ARB) and High-Pressure Sliding (HPS)

Yongpeng Tang; Toshiki Fujii; Shoichi Hirosawa; Kenji Matsuda; Daisuke Terada; Zenji Horita

doi:10.2320/matertrans.MT-MF2022047

Abstract

This study presents that A1050 commercial-purity aluminum increases the tensile strength and ductility using the processes of accumulative roll bonding (ARB) and high-pressure sliding (HPS). Both processes yield a similar tensile strength exceeding 240 MPa after processing by ARB for 10 cycles and by HPS for the sliding distance of 15 mm, respectively. The stress-strain behavior is evaluated through microstructure observations and measurements of strain hardening rates. Significant grain refinement with well-defined grain boundaries is responsible for the strength increase. The grain refinement also leads to an increase in strain hardening rate and thus an increase in the ductility.

Fig. 2 Nominal stress-strain curves at strain rate of 1 × 10⁻⁵ s⁻¹ for 0, 1, 3, 5 and 10 cycled ARB samples.

1. Introduction

The process of severe plastic deformation (SPD) produces ultrafine-grained (UFG) structures in many metallic materials, where the grain size is reduced to the range of less than 1 µm.¹^–¹¹⁾ It was shown that the UFG structures provide superior high strength but relatively low ductility, when compared with the coarse-grained (CG) counterparts.¹²^–¹⁵⁾ It is thus challenging to produce metallic materials with balanced combinations of high strength and high ductility because the strength and ductility usually exhibit an inverse relationship,¹⁶⁾ which is often called “Paradox of Strength and Ductility”.¹⁷⁾ Several strategies were proposed to achieve both high strength and high ductility in UFG metals.¹⁸^–²¹⁾ However, such strategies are applicable only under specific conditions: fine precipitation in age-hardenable alloys,²²^–²⁶⁾ high densities of nanotwins,²⁷^,²⁸⁾ and fabrication of bimodal grain structures.²⁹^–³³⁾

Nevertheless, Valiev et al. reported that an increase in strength was accompanied by high ductility in pure copper processed by SPD through equal-channel angular pressing (ECAP) and in pure Ti through high-pressure torsion (HPT).¹⁷⁾ Höppel et al. found that enhanced strength and ductility were attained in UFG aluminum with a commercial purity (99.5%Al) by processing using accumulative roll bonding (ARB) and ECAP.³⁴^,³⁵⁾ Höppel et al. considered that the enhanced ductility is due to an increase in strain rate sensitivity but commented that the value seemed sufficiently small to account for.³⁴⁾ Hashemi and coworkers investigated the formability and fracture toughness in the same purity Al after processing by ARB including the effect of the processing temperatures. They reported that both the formability and fracture toughness are increased by increasing numbers of ARB cycles.³⁶^–³⁸⁾ Kamikawa et al.³⁹^,⁴⁰⁾ analyzed a reason for the strengthening of pure Al after ARB processing and showed that extra strengthening occurs over the one expected from the Hall-Petch relation.⁴¹^,⁴²⁾ Gashti et al.⁴³⁾ reported the effect of grain size on strain hardening behavior in ARB-processed Al A1050. Most recently, Adachi et al.⁴⁴⁾ conducted comprehensive measurements of dislocation density through in situ XRD analysis during tensile deformation.

According to recent reviews of SPD processes,⁴⁵^–⁴⁷⁾ the ARB process and high-pressure sliding (HPS) process are the most promising methods for upsizing the sheet sample.⁴⁸^–⁵²⁾ The HPS process shares a feature similar to the HPT process due to the operation under high pressure, and it is possible to apply not only for the sheet form⁵³^,⁵⁴⁾ but also for the rod form⁵⁵^–⁵⁷⁾ and the pipe form.⁵⁸⁾ Furthermore, the HPS process provides an important advantage that the area size of sheet sample can be increased with combination of a feeding process, as called incremental feeding HPS (IF-HPS).⁵⁹^,⁶⁰⁾ However, to the best of the authors’ knowledge, no comparison has been made between the results of the ARB and HPS processes. Therefore, this study aims to investigate the effect of the straining process on mechanical properties including not only of tensile strength and ductility but also of strain hardening rate and strain rate sensitivity in an A1050 aluminum. Comparison is also made for microstructures observed by transmission electron microscopy (TEM).

2. Experimental

This study used aluminum sheets with 1 mm thickness in a purity level of 99.5%Al (commercially designated A1050). The sheet was then annealed at 623 K for 1.8 ks in air, which led to a mean grain size of 31 µm measured using optical microscopy (OM) as shown in Fig. 1. For the ARB processing, two pieces of the annealed strips with 75 mm width and 120 mm length were stacked and roll-bonded at room temperature by 50% reduction in one pass without using lubricant. Here, one cycle of ARB processing is defined as a procedure involving cutting, stacking and roll-bonding, and this process was repeated to 1, 3, 5 and 10 cycles. (In this study, the annealed sample is denoted as 0-cycled.) Note that the rolling direction (RD) is simplex in all the ARB cycles.

Fig. 1

OM observation of original microstructure.

To compare the processing effect on the mechanical properties, strips having dimension of 10 mm in width and 100 mm in length were also subjected to the HPS processing. The sliding process was made under a pressure of 2 GPa at a sliding speed of 1.0 mm/s for a sliding distance of X = 15 mm at room temperature (R.T.).

Tensile specimens with a parallel gauge section in 30 mm length, 7 mm width and 1 mm thickness were cut from the sheets for ARB and HPS samples. For the ARB-processed samples, the specimens were cut along the rolling direction, while for the HPS-processed samples, the specimens were cut along the sliding direction. Uniaxial tensile tests were carried out at room temperature using a tensile machine (A&D RTF-1350) at a selected strain rate out of 1 × 10⁻², 1 × 10⁻³, 1 × 10⁻⁴ and 1 × 10⁻⁵ s⁻¹.

Microstructures were observed by transmission electron microscopy (TEM). For the ARB-processed samples, TEM observation was conducted on the longitudinal cross-section parallel to the rolling direction (RD plane) as well as the normal direction (ND plane). For the HPS-processed samples, the TEM observation was carried out on the longitudinal cross-section parallel to the sliding direction as well as the normal direction. Electron-transparent thin specimens for TEM observation were prepared by a twin-jet electro-polishing technique with 15 vol% nitric acid in 85 vol% methanol at ∼253 K. TEM microstructures were observed using a transmission electron microscope (JEOL JEM 2100F) at an accelerating voltage of 200 kV.

Dislocation density measurements were carried out by X-ray diffraction (XRD) analyses using CuKα radiation with a scanning speed of 0.2°/min and a scanning step of 0.01°. Dislocation density were estimated from peak broadening using the Williamson-Hall method.⁶¹^,⁶²⁾ The calculation was carried out using five different full widths at half maximum (FWHMs) obtained from (111), (200), (220), (311) and (222) planes after normalization by calibrating the instrument.⁶³⁾

3. Results

3.1 Mechanical properties

Figure 2 shows the nominal stress-strain curves of the 0, 1, 3, 5, 10-cycled ARB samples at a strain rate of 1 × 10⁻⁵ s⁻¹. The 0-cycled sample exhibits a tensile behavior typical of coarse-grained aluminum with a yield strength (σ_y) of ∼40 MPa and an ultimate tensile strength (σ_uts) of ∼80 MPa accompanied by large strain hardening leading to an elongation to failure as large as ∼46%. By contrast, the ARB-processed samples significantly increased the strength in terms of σ_y and σ_uts but markedly decreased the elongation to failure. Nevertheless, simultaneous enhancement of the strength and ductility was observed with the increasing number of ARB cycles at a strain rate of 1 × 10⁻⁵ s⁻¹. The values of σ_y and σ_uts increased from 106 and 127 MPa for the 1-cycled sample to 128 and 178 MPa for the 10-cycled sample, respectively, with a total elongation increased from 6% to 21%. It should be noted that this is contrary to the conventional mechanical behavior of Al alloys where the strength increases at the expense of the ductility.¹⁵⁾ However, the present result is consistent with the mechanical behavior of pure Al subjected to ARB processing where both strength and ductility increase.³⁴^,³⁵⁾

Fig. 2

Nominal stress-strain curves at strain rate of 1 × 10⁻⁵ s⁻¹ for 0, 1, 3, 5 and 10 cycled ARB samples.

Figure 3 illustrates nominal stress-strain curves with a selected strain rate in the range from 1 × 10⁻² to 1 × 10⁻⁵ s⁻¹ for (a) 0, (b) 3, (c) 5 and (d) 10-cycled ARB samples. The stress-strain curve is insensitive to the strain rate in the 0-cycled samples as shown in Fig. 3(a), whereas it is sensitive in the 3, 5 and 10-cycled samples as shown in Fig. 3(b), (c), (d). The strain rate dependence of the nominal stress-strain curve is more prominent as the number of the ARB cycle increases. The elongation to failure significantly increases with decreasing the strain rate. For instance, for the 10-cycled samples in Fig. 3(d), as the strain rate increases from 1 × 10⁻⁵ to 1 × 10⁻² s⁻¹, the values of σ_y and σ_uts increase from 128 MPa and 178 MPa to 197 MPa and 241 MPa, respectively, while the total elongation decreases from 21% to 5%.

Fig. 3

Nominal stress-strain curves with strain rates ranging from 1 × 10⁻² to 1 × 10⁻⁵ s⁻¹ for ARB (a) 0, (b) 3, (c) 5 and (d) 10-cycled samples.

Figure 4 shows nominal stress-strain curves after processing by HPS. The tensile tests were carried out with the strain rates in the same range as for the ARB-processed samples. Inspection shows that the strain rate dependence of the nominal stress-strain curves for the HPS-processed samples is similar to those for the ARB 10-cycled samples delineated in Fig. 3(d). For further comparison, the values of σ_y, σ_uts and E_f measured from Fig. 3 and Fig. 4 are plotted in Fig. 5 as a function of equivalent strain (ε_eq). The values by the HPS process are well comparable to those by the ARB process, although the value of σ_y appears to be higher.

Fig. 4

Nominal stress-strain curves with strain rates ranging from 1 × 10⁻² to 1 × 10⁻⁵ s⁻¹ for HPS processed samples.

Fig. 5

Yield stress and elongation to failure plotted against equivalent strain at strain rate of 1 × 10⁻⁵ s⁻¹.

Here, the equivalent strain (ε_eq) was estimated using the following equations (1) and (2) for the ARB process⁴⁹⁾ and the HPS process,⁵⁾ respectively:

\begin{equation} \varepsilon_{\text{eq}} = \frac{2N}{\sqrt{3}}\ln(2) = 0.8N \end{equation}

(1)

where N is the number of ARB cycles.

\begin{equation} \varepsilon_{\text{eq}} = \frac{x}{\sqrt{3}t} \end{equation}

(2)

where, x is the sliding distance and t is the thickness of sample. In this study, x = 15 mm and t = 1.0 mm were used for the calculation of ε_eq which results in 8.7 and is similar to 8.0 after 10 cycles by ARB.

The strain rate dependence of the flow stress may be evaluated more quantitatively using the strain rate sensitivity m defined as

\begin{equation} m = (\partial \ln \sigma/\partial \ln \dot{\varepsilon})_{\varepsilon} \end{equation}

(3)

where σ is the flow stress and $\dot{\varepsilon }$ is the strain rate.⁶⁴⁾ Figure 6(a) plots the relationship between ln σ and $\ln \dot{\varepsilon }$ at a stain of 0.002 (= 0.2%) where the slope corresponds to m. Whereas m = 0.001 for the 0-cycled sample, the value of m increases as the number of the ARB cycle increases and it takes 0.063 after 10 cycles. The increase in m is also well demonstrated in Fig. 6(b). It is noted that the value of m increases to 0.058 after HPS processing through the sliding distance of x = 15 mm. Such an enhanced strain rate sensitivity agrees with earlier reports where commercial purity Al was subjected to SPD processing.³⁴^,³⁵^,⁶⁵^,⁶⁶⁾

Fig. 6

(a) Double logarithmic plots of flow stress and strain rate at strain of 0.002 for ARB 0, 1, 3, 5 and 10-cycled samples and HPS processed samples, where slope corresponds to strain rate sensitivity, and (b) plots of strain rate sensitivity against number of ARB cycles, including the value of HPS processed sample.

Figure 7 is typical TEM microstructures after ARB processing. Elongated grains are well developed after 5 (Fig. 7(a)) and 10 (Fig. 7(b)) cycles along the rolling direction with mean grain thicknesses of 310 and 240 nm as shown in Fig. 8, respectively, where the corresponding aspect ratio was measured to be 5.6 and 8.3. Therefore, when compared with the mean grain size after annealing which is 31 µm, significant grain refinement was achieved by the ARB processing. It should be noted that the grain sizes measured in this study are reasonably consistent with 340 nm after 7 cycles and 270 nm after 9 cycles reported by Gashti et al.⁴³⁾ on the same commercial pure Al (A1050).

Fig. 7

TEM bright-field images of ARB (a) 5-cycled and (b) 10-cycled samples.

Fig. 8

Distribution frequency of grain size after (a) ARB and (b) HPS processings.

Figure 9 shows TEM bright-field images with selected-area electron diffraction (SAED) patterns after HPS processing with a sliding distance of 15 mm. Equiaxed grains with an average grain size of ∼450 nm were observed (Fig. 8(b)), which is larger than the thickness of the elongated grains after 10 cycles of ARB processing. The increase in grain size is attributed to the gradual formation of high-angle grain boundaries under high pressure, as previously reported.⁵^,⁶⁷⁾ However, the resulting grain size is still smaller than the earlier reported for A1050 commercial-purity Al processed by HPT.⁶⁷⁾ It is noted that dislocations have been observed within the interior of UFGs, as indicated by arrows. These dislocations are believed to contribute to the enhanced strength, relative to the yield strength of the sample processed by ARB 10 cycles.

Fig. 9

TEM bright-field images of HPS processed samples.

TEM observation was conducted after tensile testing at 1 × 10⁻⁵ s⁻¹ for ARB 0-cycled and 5-cycled samples, and typical microstructures are shown with a form of montage in Fig. 10(a) and (b), respectively. It should be noted that the tensile deformation was terminated for the microstructure observation after straining to 0.4 for the former sample and to 0.1 for the latter sample. Dislocation cells are visible in the 0-cycled sample, which consist of dislocation walls and the areas surrounded by the walls. The dislocation walls appear to be agglomeration of a high density of dislocations, while a few dislocations are present between the walls. By contrast, the microstructural feature is quite different in the 5-cycled sample. There are no dislocation walls but well-defined grain boundaries. Some dislocations are visible between the grain boundaries as shown in Fig. 10(b).

Fig. 10

TEM bright-field images after tensile testing at 1 × 10⁻⁵ s⁻¹ to (a) total strain of 0.4 for ARB 0-cycled (annealed) sample and (b) total strain of 0.1 for 5-cycled sample.

4. Discussion

4.1 Strengthening by UFG structure

As shown in Fig. 2, the tensile strength increased with increasing the ARB cycles. It is suggested that the enhanced strength should be mainly due to the grain refinement and the accumulation of dislocations because neither precipitation nor twin formation is expected in the present pure aluminum. The yield stress, σ_y, may be estimated as follows:

\begin{equation} \sigma_{y} = \sigma_{0} + \Delta \sigma_{\rho} + \Delta \sigma_{\text{GB}} \end{equation}

(4)

where σ₀ is the friction stress, and Δσ_ρ and Δσ_GB are the strengths contributed from dislocations and grain boundaries, respectively, which may be estimated from the following Bailey-Hirsch⁶⁸⁾ and Hall-Patch relationships:⁴¹^,⁴²^,⁶⁹⁾

\begin{equation} \Delta \sigma_{\rho} = \alpha MGb\sqrt{\rho} \end{equation}

(5)

\begin{equation} \Delta \sigma_{\text{GB}} = \frac{k}{\sqrt{d}} \end{equation}

(6)

where α is a constant, M is the Taylor factor, G is the shear modulus and b is the Burgers vector, ρ is the dislocation density, k is the Hall-Petch coefficient and d is the average grain size. For Δσ_ρ, we use α = 0.33, M = 3.06, G = 26 GPa and b = 0.286 nm, and ρ is can be calculated from the HV taken from the reports by Gashti et al.⁴³⁾ so that ρ = 1.8 × 10¹³ m⁻² for 5 cycles and ρ = 1.5 × 10¹³ m⁻² for 10 cycles. Their dislocation densities were used for this calculation because the flow stress level is very similar to the one in this study. The dislocation density processing by HPS is measured to 5 × 10¹³ m⁻². The calculation through eq. (5) then yields Δσ_ρ = 32 MPa, 29 MPa and 53 MPa, respectively. For Δσ_GB, we use k = 0.04 MPa/m^1/2 from a report by Hansen.⁷⁰^,⁷¹⁾ With increasing the ARB cycles from 5 to 10 cycles, the thickness of elongated grain was decreased from 310 nm to 240 nm as shown in Fig. 7 and Fig. 8. Consequently, the calculated values of Δσ_GB for 5 and 10 cycles are 72 and 82 MPa, respectively. Here, the grain thickness was used as it is approximately equal to the size determined by the linear intercept method. The use of the grain thickness can be justified as described in Appendix. The values of Δσ_GB is 60 MPa in the HPS-processed sample.

Finally, the value of σ₀ may be estimated to be 8 MPa from the coarse-grained annealed sample through eq. (4) such that σ_y = 31 MPa is subtracted by the contribution of dislocations and grain boundaries, Δσ_ρ = 16 MPa and Δσ_GB = 7 MPa, respectively. Here, the dislocation density at the annealed state is measured to 5 × 10¹² m⁻² and the grain size, 31 µm, is used from the measurement in this study.

Figure 11 compares between the yield stresses measured from Fig. 3 at the given strain rate of 1 × 10⁻⁵ s⁻¹ and the ones estimated through eq. (2). It appears that the calculated values are invariably lower than the experimental values. This difference is reasonable because extra hardening occurs due to the stress required for the generation of mobile dislocations within small grains as pointed out by Huang et al.⁵⁰⁾ The values obtained from the HPS-processed samples also shows higher hardness than the ones expected from the Hall-Petch relation.

Fig. 11

Plots of yield stress against number of ARB cycle at strain rate of 1 × 10⁻⁵ s⁻¹, including value of HPS processed sample. Closed symbols with solid line for yield stresses measured from Fig. 3 and Fig. 4, while open symbols with dotted line for yield stresses estimated using eq. (2).

4.2 Ductility of UFG structure

In this study, we define the ductility as the extent to which the material can plastically deform to failure. This study thus demonstrated that the concurrent enhancement of strength and ductility was achieved with increasing the number of ARB cycles, particularly when the strain rate was low. The results are essentially in agreement with the reports by Höppel et al.³⁴⁾ The ductility may be evaluated in terms of uniform and local deformation, where the uniform deformation is up to the onset of necking which is followed by the local deformation with plastic instability.

The following condition was derived by Hart for the stable (uniform) deformation before necking.⁷²⁾

\begin{equation} \gamma + m \geq 1 \end{equation}

(7)

where γ is given as follows and m is defined in eq. (3) as above:

\begin{equation} \gamma = (\partial\sigma/\partial\varepsilon)_{\dot{\varepsilon}}/\sigma \end{equation}

(8)

Thus, to increase the uniform deformation before necking, it is important that both the strain hardening rate, $(\partial \sigma /\partial \varepsilon )_{{\dot{\varepsilon }}}$, and the strain rate sensitivity, m, should be high. Hart commented that, when $m \simeq 0$ as observed in the 0-cycled sample, it reduces to the Considère criterion for uniform deformation as

\begin{equation} (\partial\sigma/\partial\varepsilon)_{\dot{\varepsilon}} \geq \sigma \end{equation}

(9)

However, when m is appreciable, the Considère criterion is modified as follows.

\begin{equation} (\partial\sigma/\partial\varepsilon)_{\dot{\varepsilon}} \geq \sigma(1-m) \end{equation}

(10)

The strain hardening rate from the Considère criterion is plotted in Fig. 12 together with the stress-strain curves. If the stress-strain curve is represented by the Hollomon equation as⁷³⁾

\begin{equation} \sigma = K\varepsilon^{n} \end{equation}

(11)

the strain hardening rate is given as

\begin{equation} (\partial \sigma/\partial \varepsilon)_{\dot{\varepsilon}} = n(\sigma/\varepsilon). \end{equation}

(12)

Where K is a constant and n is the strain hardening coefficient. Thus, the intersection between the two curves in Fig. 12, which is equivalent to $(\partial \sigma /\partial \varepsilon )_{{\dot{\varepsilon }}} = \sigma $, gives rise to

\begin{equation} n = \varepsilon \end{equation}

(13)

This indicates that the strain hardening coefficient is equal to the strain corresponding to the uniform deformation before necking.

Fig. 12

True stress and strain-hardening rate plotted against true strain after tensile testing at strain rate of 1 × 10⁻⁵ s⁻¹ for (a) 0, 1, 3, 5 and 10-cycled samples and HPS-processed sample. (b) Enlargement for lower strain range.

The values of n so determined are plotted in Fig. 13 as a function of the number of the ARB cycle, including the HPS processed sample. The values of n obtained from the Hollomon equation are also determined from the slope in a plot of ln σ vs. ln ε as shown in Fig. 14. For this determination, the straight portions were used. Such n values are also plotted in Fig. 13 and compared with the ones determined from the intersections in Fig. 12. Both types of n values show almost the same tendency. Although the n value decreases significantly by the first cycle, it increases with the number of the ARB cycle. After completing ARB 10 cycles, the n value is consistent with the value obtained from the HPS-processed samples. Close comparison reveals that the values determined from the slopes in Fig. 14 tend to be invariably lower. It is considered that this discrepancy must be due to the use of eq. (9) which is effective when $m \simeq 0$. However, when m is appreciable as shown in Fig. 6, the flow stress level should be lower by (1 − m), and this leads to an increase in n (= ε). Thus for this study, the effect of m increases the discrepancy. Nevertheless, because the m values determined in this study are not high as shown in Fig. 6 (i.e. the highest is m = 0.063), which is consistent with other published reports,³⁴^,³⁵^,³⁹^,⁶⁶^–⁶⁹^,⁷²^–⁷⁴⁾ the related shift for n is trivial.

Fig. 13

Plots of strain hardening coefficient against ARB cycle, including value of HPS-processed sample.

Fig. 14

Double logarithmic plots of flow stress and strain for 0, 3, 5 and 10-cycled samples and HPS-processed sample, where slope corresponds to strain hardening coefficient, n, after tensile deformation at strain rate of 1 × 10⁻⁵ s⁻¹.

Another possible reason for the discrepancy may be due to the use of the form of eq. (11) given by Hollomon. Alternatively, the following Ludwik’s equation may be used because the stress-strain curve does not fit for an early stage of strain.⁷⁴⁾

\begin{equation} \sigma = \sigma_{y} + K\varepsilon^{n} \end{equation}

(14)

where σ_y is the yield stress as defined earlier. As Fig. 14, the plot of ln(σ − σ_y) vs. ln ε leads to n from the slope. The values of n so determined are also included in Fig. 13 and they are now slightly higher than those determined from the intersection through $(\partial \sigma /\partial \varepsilon )_{{\dot{\varepsilon }}} = \sigma $. It should be noted that the intersection now corresponds to the following condition when eq. (14) (Ludwik’s equation) is used:

\begin{equation} (\partial \sigma/\partial \varepsilon)_{\dot{\varepsilon}} \geq \sigma - \sigma_{y} \end{equation}

(15)

Thus, the values of n determined through $(\partial \sigma /\partial \varepsilon )_{{\dot{\varepsilon }}} = \sigma - \sigma_{y}$ are higher than those determined using $(\partial \sigma /\partial \varepsilon )_{{\dot{\varepsilon }}} = \sigma $.

4.3 Simultaneous increase in strength and ductility

Figure 15 displays the uniform and local deformations together with the tensile strength measured from the stress-strain curves after 0, 1, 3, 5 and 10-cycles shown in Fig. 3 and the HPS-processed samples in Fig. 4. Here, the uniform deformation was determined from the intersection in Fig. 12. It is seen that the uniform deformation markedly decreased in the ARB-processed samples when compared with the unprocessed sample (i.e., 0-cycled sample). However, the uniform deformation tends to increase with the increasing number of ARB cycle. Furthermore, the local deformation also increases with the increasing number of ARB processing. Therefore, the ARB process results in not only an increase in the tensile strength but also in ductility by increasing the cycle number. It should be noted that the HPS-processed sample exhibits a similar elongation to the sample processed by ARB 10 cycles.

Fig. 15

Histogram showing variation of total elongation consisting of uniform and local deformations and plot showing variation of tensile strength for 0, 1, 3, 5 and 10-cycles and HPS-processed sample after tensile testing at 1 × 10⁻⁵ s⁻¹.

To consider a reason for the ductility increase, TEM observations shown in Fig. 7 and Fig. 10 may provide an insight. Well defined grain boundaries were developed after processing for 5 and 10 cycles as shown in Fig. 7, and the grain size is smaller after 10 cycles than 5 cycles as given in Fig. 8. Tensile deformation reveals, as shown in Fig. 10, that dislocation cell walls were not visible in the 5-cycled sample but they were in the annealed sample. Yu et al. reported⁷⁵⁾ that strain hardening is improved when the formation of dislocation cells is inhibited. The pillar size smaller than the dislocation cell size exhibited a large increase in strain hardening rate, and thus they suggested that a similar enhancement of the strain hardening occurs if the grain size is smaller than the dislocation cell size. As discussed in association with the Hart’s equation given in eq. (7), the uniform elongation increases as the strain hardening rate increases. The high-angle grain boundaries acted as a dislocation sink so that they played a role similar to the sample surface where dislocations disappear. It should be also commented that the dislocations disappeared at the grain boundaries and this produces the higher ductility including the local deformation after necking. Although there is no direct evidence for the absorption of dislocations at grain boundaries, it is suggested from the measurements of dislocation density by Miyajima et al.⁴⁴⁾ They reported that the dislocation density significantly increases after the first cycle and takes a maximum after the second cycle. Thereafter, the dislocation density decreases gradually and this decrease is consistent with the decrease in grain boundary spacing. According to an in situ measurement of dislocation density during tensile testing by Adachi et al.,⁴⁵⁾ it decreases gradually after reaching a stress maximum and decreases suddenly due to unloading by fracture in a 6-cycled sample. This is contrast to the case of a coarse-grained sample: the dislocation density increases without the stress maximum before the fracture where the decrease in the dislocation density is much less than the ARB-processed sample. Thus, the dislocation behavior is well different between the two cases and is considered to be attributed to the absorption at grain boundaries.

5. Summary and Conclusions

(1) A1050 commercial-purity aluminum was processed for grain refinement by ARB and HPS. Both processes refined the grain size to 240 nm and 450 nm after processing by the ARB 10 cycles and by the sliding distance of 15 mm, respectively. Strain hardening rate and strain rate sensitivity were measured from the stress-strain curves for the determination of uniform elongation.
(2) Processing by both ARB and HPS led to a marked increase in the tensile strength (σ_uts), reaching 240 MPa and 243 MPa after processing by the ARB 10 cycles and by HPS for the sliding distance of 15 mm, respectively. The strain rate sensitivity was also found in both processed samples during the tensile deformation process. The increased strain rate sensitivity value has played a significant role for the improvement of the ductility.
(3) Not only uniform elongation but also local deformation was increased as the number of ARB cycle increases. This increase is attributed to an increase in strain hardening rate due to absorption of dislocations at the well-defined grain boundaries.
(4) Tensile strength as well as the yield stress was significantly enhanced as the number of ARB cycle increases. This increase in the strength was consistent with the strength predicted by the Bailey Hirsch equation and the Hall-Petch relation plus extra hardening due to the stress required for the generation of mobile dislocations within small grains.

Acknowledgments

The authors would like to express their gratitude to Dr. Yoichi Takizawa of Nagano Forging Co., Ltd. for his assistance in processing the HPS-processed samples. The authors are grateful to Japan Science and Technology Agency (JST) under collaborative research based on industrial demand “Heterogeneous structure control: Towards innovative development of metallic structural materials” to support this study. In addition, this work was supported in part by the Light Metals Educational Foundation of Japan and in part by a Grant-in-Aid for Scientific Research (A) from the MEXT, Japan (JP19H00830).

REFERENCES

Appendix

According to Smith and Guttman,⁷⁶⁾ the following relation holds

\begin{equation} \text{L} = 2/\text{S}_{\text{V}} \end{equation}

(A.1)

where L is the average number of intercepts per unit length and S_V is the surface area per unit volume. Thus, S_V is given as

\begin{equation} \text{S}_{\text{V}} = (\text{d}_{w}\text{d}_{t} + \text{d}_{t}\text{d}_{l} + \text{d}_{l}\text{d}_{w})/\text{d}_{w}\text{d}_{t}\text{d}_{l} \end{equation}

(A.2)

Here, for calculation, the shape of the grain is approximated to be a plate with dimensions of the thickness d_t, the width d_w and the length d_l.

Assuming that the grain is equally enlarged to both the rolling and transverse directions during ARB process and thus $\text{d}_{w} \simeq \text{d}_{l}$,

\begin{equation} \text{S}_{\text{V}} = (\text{2d}_{t}/\text{d}_{l} + 1)/\text{d}_{t} \end{equation}

(A.3)

We now consider the two extreme cases: one is $\text{d}_{t}/\text{d}_{l} \simeq 0$, where the grain is largely elongated (i.e. the aspect ratio is very large) and the other is $\text{d}_{t}/\text{d}_{l} \simeq 1$, where the grain is equiaxed. Thus from eq. (A.1) and eq. (A.3), it follows that

\begin{equation*} \text{0.5L} < \text{d}_{t} < \text{1.5L} \end{equation*}

Considering that the present cases where the aspect ratios are 5.6 and 8.3 after 5 and 10 cycles, respectively, it may be reasonable that $\text{d}_{t} \simeq \text{L}$. This suggests that we can use the grain thickness for the size determined by the conventional linear intercept method.

Corresponding author

Version information

Register with J-STAGE for free!